At BBVA, nine lawyers support Spanish branch managers who bring them more than 40,000 customer legal questions a year. Answering can require searching product documents and internal and external rules. In January 2025, the bank described an assistant its legal team had built in ChatGPT to consult those materials and prepare responses.
Lawyers reviewed every response before it reached a branch manager. BBVA said the system helped the team answer more questions within 24 hours. The bank gave no before-and-after count in that announcement, so the size of that improvement remained unspecified.
Another BBVA assistant helped employees navigate its central store of business data. Staff could ask where to find information, who owned it and which other applications had used it.
By May 2025, BBVA was expanding from 3,300 to 11,000 ChatGPT Enterprise licenses. According to its AI-adoption department, 83 percent of licensed users were using the service daily. Employees who automated tasks estimated they saved an average of 2.8 hours a week.
The bank had created an internal library to help employees find assistants other teams had already built. About 1,000 were considered useful across the organization. Training, regular meetings and online discussions accompanied the software rollout, giving employees places to exchange examples and ask colleagues for help.
An adviser edits before sending
Morgan Stanley’s Debrief tool, announced in June 2024, worked at a particular point in an adviser’s day: after a client meeting. With client consent, it prepared notes and action items, drafted a follow-up email and saved a note in Salesforce, the firm’s customer-management software. The adviser could edit the email and decide whether to send it.
Don Whitehead, an adviser in Houston, said in the company’s announcement that the tool saved him about half an hour per meeting. Zach Goldberg, an adviser in Short Hills, New Jersey, described reviewing and sending a summary while the conversation was still fresh. Both accounts were testimonials supplied by Morgan Stanley.
Morgan Stanley also reported adoption of a separate knowledge-search Assistant by 98 percent of adviser teams, without an individual daily-use figure.
Customer service produces a different set of figures
Klarna put an AI assistant directly in front of customers seeking help with shopping and payments. Its 2025 annual report said the assistant handled 80 percent of customer-service chats during that year and delivered approximately $59 million in cost savings. Klarna estimated the workload as equivalent to more than 850 full-time agents, based on reduced chat and telephone volumes handled by staff.
The company reported similar satisfaction levels for AI and human service, citing its own surveys. It also said all customers retained the option to speak to a person.
Klarna was using different systems for different jobs. Its annual report described AI software that classified more than a million chat conversations a month and an internal assistant called Kiki that helped staff find company information. It separately said credit underwriting used established machine-learning methods, while generative AI was not used to make those lending decisions.
.jpg){kind=link}